I think we're teaching the wrong things for an AI world

I've been spending quite a lot of time thinking about robotics and AI in education recently. Some of that has come from conversations about new learning environments and the technology that might go into them, but it's increasingly made me think about a much bigger question. If AI and robotics continue to develop at anything like their current pace, what exactly are we preparing young people for?

Most conversations about this seem to start with technology. How do we teach more AI? Should students learn robotics? What technical skills will employers need? What equipment should colleges invest in? These are all sensible questions and, given the speed of change, education clearly needs to respond. I'm just not convinced they're the best place to start.

There's something slightly odd about looking at technologies that are becoming extraordinarily good at technical and cognitive tasks and concluding that the answer is simply to teach people more technical tasks. Of course people need to understand the technology. I'd argue that a basic understanding of AI will become as normal as understanding the internet is today. But knowing how to use technology and deciding what to do with it are quite different things.

The latter is where this gets interesting for me. As machines become more capable, some very human abilities may become more valuable rather than less: judgement, curiosity, creativity, communication, empathy and the ability to understand context. We often bundle these together under the rather dismissive label of “soft skills”, which I've never particularly liked. There's nothing soft about making a good decision when there isn't an objectively correct answer, persuading people with competing interests to work together, or recognising that the technically optimal solution might be completely wrong for the people affected by it.

Robotics makes this particularly obvious. We're now seeing machines that can navigate physical environments, interpret what they see and perform increasingly sophisticated tasks. Inevitably, much of the education conversation is about teaching students to programme, operate and maintain them. We should absolutely do that. But somebody also has to decide what those machines should be allowed to do, how they should behave around people and who is responsible when something goes wrong.

Those aren't simply engineering problems. They're questions involving ethics, law, psychology, sociology, philosophy and design. A robotics student who understands Python, computer vision and control systems is clearly valuable. A robotics student who understands those things but can also think seriously about the social and ethical consequences of putting an autonomous machine into a hospital, factory, school or public space may be considerably more valuable.

This makes me wonder whether some of the disciplines we've tended to regard as being slightly removed from the “future skills” conversation are about to become much more important. There has understandably been a huge push towards STEM over the past couple of decades, but perhaps AI makes the old distinction between technical education and the humanities less useful. The interesting problems increasingly sit between them.

There's a practical issue here too. Technical knowledge has always dated, but it now seems to be dating extraordinarily quickly. Someone can spend years becoming proficient in a particular tool or workflow only for AI to automate a significant part of it. That doesn't make technical education pointless; far from it. It does, however, make me wary of defining employability around proficiency in whatever happens to be today's technology.

Perhaps the more durable skill is being able to encounter a technology you haven't seen before, understand enough about it to use it intelligently, work out where it is genuinely useful and recognise where its limitations lie. That requires technical confidence, but it also requires curiosity, judgement and critical thinking. In other words, the technology changes but some of the underlying human capabilities travel remarkably well.

It also has implications for the learning environments we're creating. If education is mainly about transferring technical knowledge, then the obvious response is to put better technology into classrooms. But if students need to learn how to experiment with technology, collaborate with other disciplines, solve ambiguous problems and work alongside increasingly intelligent machines, the environment has to support something rather different. It needs to encourage experimentation, discussion, making things, getting things wrong and trying again. It should make it easier to bring together students, educators, employers and specialists who don't necessarily come from the same discipline.

That's much more interesting to me than starting with a shopping list of equipment. A robot, an immersive room or an AI platform might be part of the answer, but none of them tells you what the learning experience should actually be.

We spend an enormous amount of time trying to predict which jobs AI will replace and which new ones it will create. I'm increasingly sceptical that anybody can do that with much confidence beyond the relatively near term. The technology is changing too quickly, and predictions about the future of work have a fairly patchy history anyway.

Perhaps education doesn't need to predict the future quite so precisely. If we can develop people who are comfortable with technology but are also curious, adaptable, creative, thoughtful and capable of exercising judgement, we're giving them something more useful than training them for our current best guess at the jobs of 2035.

For years, the debate has often been framed as a need for more STEM. I wonder whether the next stage is actually about bringing STEM and the humanities much closer together. The more capable our machines become, the less interesting it may be to ask what humans can do that machines can't.

The better question might be what we want humans to be good at.